DEEPre: sequence-based enzyme EC number prediction by deep learning

• DOI: 10.1093/bioinformatics/btx680
• Publication Date: 2017-10-20T19:13:35Z
• Authors: Yu Li; Sheng Wang; Ramzan Umarov; Bingqing Xie; Ming Fan; Lihua Li; Xin Gao
• Keywords: UniProt

Annotation of enzyme function has a broad range applications, such as metagenomics, industrial biotechnology, and diagnosis deficiency-caused diseases. However, the time resource required make it prohibitively expensive to experimentally determine every enzyme. Therefore, computational prediction become increasingly important. In this paper, we develop an approach, determining by predicting Enzyme Commission number.We propose end-to-end feature selection classification model training well automatic robust dimensionality uniformization method, DEEPre, in field prediction. Instead extracting manually crafted features from sequences, our takes raw sequence encoding inputs, convolutional sequential based on result directly improve performance. The thorough cross-fold validation experiments conducted two large-scale datasets show that DEEPre improves performance over previous state-of-the-art methods. addition, server outperforms five other servers main class enzymes separate low-homology dataset. Two case studies demonstrate DEEPre's ability capture functional difference isoforms.The could be accessed freely at http://www.cbrc.kaust.edu.sa/[e-mail] data are available Bioinformatics online.